Adaptive weighting method for orthogonal frequency division multiplexed soft symbols using channel state information estimates

CROSS REFERENCE TO RELATED APPLICATION

0001This application is a continuation application of application Ser. No. 09/153,636, filed Sep. 15, 1998, now U.S. Pat. No. 6,317,470.

BACKGROUND OF THE INVENTION

0002This invention relates to signal processing, and more particularly, to signal processing techniques for use in Digital Audio Broadcasting (DAB) systems.

0003Digital Audio Broadcasting is a medium for providing digital-quality audio, superior to existing analog broadcasting formats. Both AM and FM DAB signals can be transmitted in a hybrid format where the digitally modulated signal coexists with the currently broadcast analog AM or FM signal, or in an all-digital format without an analog signal. In-band-on-channel (IBOC) DAB systems require no new spectral allocations because each DAB signal is simultaneously transmitted within the same spectral mask of an existing AM or FM channel allocation. IBOC promotes economy of spectrum while enabling broadcasters to supply digital quality audio to their present base of listeners. Several IBOC DAB approaches have been suggested. One such approach, set forth in U.S. Pat. No. 5,588,022, presents a method for simultaneously broadcasting analog and digital signals in a standard AM broadcasting channel. Using this approach, an amplitude-modulated radio frequency signal having a first frequency spectrum is broadcast. The amplitude-modulated radio frequency signal includes a first carrier modulated by an analog program signal. Simultaneously, a plurality of digitally-modulated carrier signals are broadcast within a bandwidth which encompasses the first frequency spectrum. Each digitally-modulated carrier signal is modulated by a portion of a digital program signal. A first group of the digitally-modulated carrier signals lies within the first frequency spectrum and is modulated in quadrature with the first carrier signal. Second and third groups of the digitally-modulated carrier signals lie outside of the first frequency spectrum and are modulated both in-phase and in-quadrature with the first carrier signal. Multiple carriers are employed by means of orthogonal frequency division multiplexing (OFDM) to bear the communicated information.

0004FM IBOC broadcasting systems using have been the subject of several United States patents including U.S. Pat. Nos. 5,465,396; 5,315,583; 5,278,844 and 5,278,826. In addition, a commonly assigned pending patent application for a “Method and System for Simultaneously Broadcasting and Receiving Digital and Analog Signals, by D. Kumar and B. Hunsinger, Ser. No. 08/274,140, filed Jul. 1994 discloses an FM IBOC DAB system, now U.S. Pat. No. 5,956,624.

0005The signals used in Digital Audio Broadcasting are subject to fading and noise (interference). Digital Audio Broadcasting receivers may include Viterbi decoders. Conventional implementations of soft-decision Viterbi decoders rely on constant signal and gaussian noise statistics for (near) optimum decoding. Practically these statistics should be nearly constant over the path memory of the Viterbi decoder, or the span of the interleaver, whichever is greater. An interleaver may be used to yield statistical independence of the fading statistics of the soft symbols over the path memory of the Viterbi decoder after deinterleaving. The remedy for a flat fading channel causing fluctuating signal levels with constant noise is well known. However, there exists a need for a signal processing technique that can address independently varying signal and noise levels.

SUMMARY OF THE INVENTION

0006A method for weighting orthogonal frequency division multiplexed soft symbols is provided, including the steps of receiving a plurality of sub-carriers modulated by digital information, filtering the sub-carriers to produce complex soft decision outputs, creating a first sequence of the magnitudes of the complex soft decision outputs, determining the differences between successive samples in the first sequence, creating a second sequence of the differences between successive samples in the first sequence, using the first and second sequences to determine a plurality of weights, and applying the plurality of weights to the complex soft decision outputs. The invention also encompasses receivers that incorporate the above method.

BRIEF DESCRIPTION OF THE DRAWINGS

<figref idref="DRAWINGS">FIG. 1</figref> is a diagram showing the steps in the method of this invention; and

<figref idref="DRAWINGS">FIG. 2</figref> is a functional block diagram showing the mapping and processing of bits through a digital audio broadcasting receiver which employs the method of this invention.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

0009The present invention will be described in terms of the adaptive weighting procedure for coherently detected quadrature phase shift keyed (QPSK) or binary phase shift keyed (BPSK) signals under various fading conditions assuming gaussian noise or interference statistics.

0010For single carrier, flat fading signal, constant noise or interference, the flat fading channel multiplies the k^thinterleaved symbol by a fading factor a_k, which is assumed to be constant over the symbol duration. The signal sample output of a matched filter in the receiver can be expressed as: <br /><i>y</i>_k<i>=a</i>_k<i>·x</i>_k<i>+n</i>_k  (1)<br /> where x_kis the transmitted signal component and n_kis the received noise. The appropriate weighting for that soft decision symbol at the receiver is itself a_k, or more practically the receiver's estimate â_k. Therefore the soft symbol s_kentering a deinterleaver prior to the Viterbi decoder is presented as: <br /><i>s</i>_k<i>=â</i>_k<i>·y</i>_k  (2)<br /> The symbols immediately surrounding the k^thsymbol are similarly affected by the relatively slowly varying fading. It is impossible to adequately estimate this fading factor over a single symbol because it is impossible to separate its statistics from the noise. However the fading factor can be estimated statistically using a weighted average over a set of samples with correlated fading constants, specifically the symbols that are temporally close to the k^thsymbol.

0011Let the signal term x_kbe complex whose real and imaginary components are binary valued at ±1. For reasonably large SNR, where the noise (or interference) is approximated to be zero mean average white Gaussian noise (AWGN), the fading factor might be estimated as: <maths id="MATH-US-00001" num="00001"><math overflow="scroll"><mtable><mtr><mtd><mtable><mtr><mtd><mrow><msub><mover><mi>a</mi><mo>^</mo></mover><mi>k</mi></msub><mo>=</mo><mrow><mi>E</mi><mo>⁢</mo><mrow><mo>{</mo><mrow><mo></mo><msub><mi>y</mi><mi>k</mi></msub><mo></mo></mrow><mo>}</mo></mrow></mrow></mrow></mtd></mtr><mtr><mtd><mrow><mo>=</mo><mrow><mrow><mi>E</mi><mo>⁢</mo><mrow><mo>{</mo><mrow><mo></mo><mrow><mrow><msub><mi>a</mi><mi>k</mi></msub><mo>·</mo><msub><mi>x</mi><mi>k</mi></msub></mrow><mo>+</mo><msub><mi>n</mi><mi>k</mi></msub></mrow><mo></mo></mrow><mo>}</mo></mrow></mrow><mo>≅</mo><mrow><mi>E</mi><mo>⁢</mo><mrow><mo>{</mo><msub><mi>a</mi><mi>k</mi></msub><mo>}</mo></mrow></mrow></mrow></mrow></mtd></mtr></mtable></mtd><mtd><mrow><mo>(</mo><mn>3</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0001.tif" /><br /> However, this estimate is not accurate for small SNR since the rectified noise will place a lower limit on â_k. This limitation would reduce erasure effectiveness when the signal level fades to zero. The signal level for coherent detection and Gaussian noise can be estimated using an exact expression as a function of the second and fourth moments of the magnitude of y_k(i.e. E{|y_k|²} and E{|y_k|⁴}). These moments are obtained through averaging or lowpass filtering of the squared or fourth power of the magnitude of y_k. Assuming a constant AWG noise floor, the weighting factor for the soft decision is equal to the fading factor: <br /><i>â</i>_k={square root over (2·(<i>E{|y|</i>²})²<i>−E{|y|</i>⁴})}  (4)<br /> This equation has been used to successfully estimate the signal level in a mobile satellite receiver.

0012For single carrier, flat fading signal, and flat fading interference, the difference between the constant signal and the flat fading signal appears as an interference fading factor σ_k. It is assumed that the fading of the signal and interference are approximately independent, and the interference is approximated as AWGN. The Gaussian interference assumption is made in the absence of known precise statistics of the interferer. Furthermore, this assumption becomes less objectionable when we recognize that the Viterbi decoder tends to “Gaussianize” the noise or interference summed over the weighted symbols in its path memories. The soft-decision output of the matched filter can be expressed as: <br /><i>y</i>_k<i>=a</i>_k<i>·x</i>_k+σ_k<i>·n</i>_k  (5)<br /> Dividing both sides by σ_kputs the expression in form familiar to the nonfaded interference case: <maths id="MATH-US-00002" num="00002"><math overflow="scroll"><mtable><mtr><mtd><mrow><mfrac><msub><mi>y</mi><mi>k</mi></msub><msub><mi>σ</mi><mi>k</mi></msub></mfrac><mo>=</mo><mrow><mrow><mfrac><msub><mi>a</mi><mi>k</mi></msub><msub><mi>σ</mi><mi>k</mi></msub></mfrac><mo>·</mo><msub><mi>x</mi><mi>k</mi></msub></mrow><mo>+</mo><msub><mi>n</mi><mi>k</mi></msub></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>6</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0002.tif" /><br /> Then the appropriate weight for the soft symbol is identified. The soft symbol entering the deinterleaver prior to the Viterbi decoder is presented as: <maths id="MATH-US-00003" num="00003"><math overflow="scroll"><mtable><mtr><mtd><mrow><msub><mi>s</mi><mi>k</mi></msub><mo>=</mo><mrow><mfrac><msub><mover><mi>a</mi><mo>^</mo></mover><mi>k</mi></msub><msubsup><mover><mi>σ</mi><mo>^</mo></mover><mi>k</mi><mn>2</mn></msubsup></mfrac><mo>·</mo><msub><mi>y</mi><mi>k</mi></msub></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>7</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0003.tif" /><br /> The fading factor can be estimated as in equation (4). An additional estimate of the noise variance <maths id="MATH-US-00004" num="00004"><math overflow="scroll"><msubsup><mover><mi>σ</mi><mo>^</mo></mover><mi>k</mi><mn>2</mn></msubsup></math></maths><img file="US6895060B2_D0004.tif" /><br /> is required for this weighting. This estimate can be computed as: <maths id="MATH-US-00005" num="00005"><math overflow="scroll"><mtable><mtr><mtd><mrow><msubsup><mover><mi>σ</mi><mo>^</mo></mover><mi>k</mi><mn>2</mn></msubsup><mo>=</mo><mrow><mrow><mi>E</mi><mo>⁢</mo><mrow><mo>{</mo><msup><mrow><mo></mo><msub><mi>y</mi><mi>k</mi></msub><mo></mo></mrow><mn>2</mn></msup><mo>}</mo></mrow></mrow><mo>-</mo><msubsup><mover><mi>a</mi><mo>^</mo></mover><mi>k</mi><mn>2</mn></msubsup></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>8</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0005.tif" />

0013For multiple sub-carriers (OFDM), selective fading signal, and selective fading interference, the method of the present invention can be described as follows. OFDM sub-carriers comprise a set of orthogonally spaced (in frequency) sub-carriers, each modulated synchronously at the symbol rate. QPSK sub-carrier modulation is assumed here, although other linear modulation techniques such as MPSK or QAM can be accommodated with modification. OFDM has been shown to be tolerant of multipath fading since the fading over any individual sub-carrier bandwidth is considered to be flat (flat fading requires no sub-carrier equalization). In a severe selective fading instance, a portion of the OFDM sub-carriers may be lost in a null while the remaining sub-carriers are recoverable. In contrast, wideband single carrier modulation can experience the same spectral null where significant degradation can occur over all the bits if equalization is not employed.

0014Each OFDM sub-carrier can be treated as an independent channel with flat fading signal and interference; however, this is not optimum. The correlation of the signal and noise fading across sub-carriers nearby in frequency can be exploited to improve performance. This can be accomplished through filtering the statistics across both time and frequency surrounding any particular sub-carrier. The k^thsymbol of the n^thsub-carrier is weighted as: <maths id="MATH-US-00006" num="00006"><math overflow="scroll"><mtable><mtr><mtd><mrow><msub><mi>s</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>=</mo><mrow><mfrac><msub><mover><mi>a</mi><mo>^</mo></mover><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><msubsup><mover><mi>σ</mi><mo>^</mo></mover><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow><mn>2</mn></msubsup></mfrac><mo>·</mo><msub><mi>y</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>9</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0006.tif" /><br /> where Y_k,nis a complex soft symbol resulting from QPSK demodulation of the k_thsymbol of the n^thsub-carrier. The only difference between equations (9) and (7) is the exploitation of the correlation of adjacent sub-carrier statistics in equation (9). The fading factors can be estimated after 2-dimensional smoothing (filtering across time and sub-carriers) of the following values: <maths id="MATH-US-00007" num="00007"><math overflow="scroll"><mtable><mtr><mtd><mrow><mrow><msub><mi>u</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>=</mo><mrow><mi>E</mi><mo>⁢</mo><mrow><mo>{</mo><msup><mrow><mo></mo><msub><mi>y</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo></mo></mrow><mn>2</mn></msup><mo>}</mo></mrow></mrow></mrow><mo>⁢</mo><mstyle><mtext></mtext></mstyle><mo>⁢</mo><mrow><msub><mi>v</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>=</mo><mrow><mi>E</mi><mo>⁢</mo><mrow><mo>{</mo><msup><mrow><mo></mo><msub><mi>y</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo></mo></mrow><mn>4</mn></msup><mo>}</mo></mrow></mrow></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>10</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0007.tif" /><br /> Compute the symbol weighting factor for coherently detected QPSK as: <maths id="MATH-US-00008" num="00008"><math overflow="scroll"><mtable><mtr><mtd><mrow><msub><mi>w</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>=</mo><mfrac><mroot><mrow><mrow><mn>2</mn><mo>·</mo><msubsup><mi>u</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow><mn>2</mn></msubsup></mrow><mo>-</mo><msub><mi>v</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub></mrow><mn>4</mn></mroot><mrow><msub><mi>u</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>-</mo><msqrt><mrow><mrow><mn>2</mn><mo>·</mo><msubsup><mi>u</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow><mn>2</mn></msubsup></mrow><mo>-</mo><msub><mi>v</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub></mrow></msqrt></mrow></mfrac></mrow></mtd><mtd><mrow><mo>(</mo><mn>11</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0008.tif" /><br /> where w_k,nis the weight to be used for the soft symbol prior to deinterleaving.

0015The adaptive weighting procedure for differentially detected QPSK signals under various fading conditions will now be described. These results can be extended to BPSK with minor modification.

0016The goal of the present invention is to maximize the resulting SNR of the weighted and combined symbols after differential detection into the Viterbi decoder. This minimizes the probability of bit error at the output of the Viterbi decoder. The absolute optimization of the weighting of the soft decision statistic is a function of the particular Viterbi decoder. Therefore the answer is not necessarily unique. Furthermore, since the Viterbi decoder tends to “Gaussianize” the path metrics over the path memories, then this technique tends to maximize the SNR over the path memory.

0017A similar expression for the weighting factor can be derived assuming Gaussian noise into a differential QPSK detector resulting in non-Gaussian statistics at the output. The fading factor can be computed as a function of the statistics of the output of the differential detector where we define the soft decision of the form: <br /><i>S</i>=(<i>a+n</i>₁)·(<i>a·e</i>^Jφ<i>+n</i>₂)  (12)<br /> where φ denotes the phase information imposed between a pair of adjacent symbols in the differential encoding, and n are the independent noise samples. The fading factor a of the adjacent symbols is assumed to be approximately equal. The signal to noise ratio after differential detection is easily computed to be: <maths id="MATH-US-00009" num="00009"><math overflow="scroll"><mtable><mtr><mtd><mrow><mi>SNR</mi><mo>=</mo><mfrac><msup><mi>a</mi><mn>4</mn></msup><mrow><mrow><mn>2</mn><mo>·</mo><msup><mi>a</mi><mn>2</mn></msup><mo>·</mo><msup><mi>σ</mi><mn>2</mn></msup></mrow><mo>+</mo><msup><mi>σ</mi><mn>4</mn></msup></mrow></mfrac></mrow></mtd><mtd><mrow><mo>(</mo><mn>13</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0009.tif" /><br /> The ideal weighting factor for the post-differentially detected symbols is therefore: <maths id="MATH-US-00010" num="00010"><math overflow="scroll"><mtable><mtr><mtd><mrow><mi>w</mi><mo>=</mo><mfrac><msup><mi>a</mi><mn>2</mn></msup><mrow><mrow><mn>2</mn><mo>·</mo><msup><mi>a</mi><mn>2</mn></msup><mo>·</mo><msup><mi>σ</mi><mn>2</mn></msup></mrow><mo>+</mo><msup><mi>σ</mi><mn>4</mn></msup></mrow></mfrac></mrow></mtd><mtd><mrow><mo>(</mo><mn>14</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0010.tif" /><br /> The first differential approach described here uses statistical estimates of the second and fourth moments of the differentially detected symbols to form the weighting factor. The second and fourth moments are described by the following previously known relationships. <br /><i>E{|S|</i>²}=(<i>a</i>²=σ²)²<br /><i>E{|S|</i>⁴}=(<i>a</i>⁴+4·<i>a</i>²·σ²+2·σ⁴)²  (15)<br /> Then the fading factor can be estimated as: <maths id="MATH-US-00011" num="00011"><math overflow="scroll"><mtable><mtr><mtd><mrow><msub><mover><mi>a</mi><mo>^</mo></mover><mi>k</mi></msub><mo>=</mo><mroot><mrow><mrow><mrow><mn>2</mn><mo>·</mo><mi>E</mi></mrow><mo>⁢</mo><mrow><mo>{</mo><msup><mrow><mo></mo><msub><mi>S</mi><mi>k</mi></msub><mo></mo></mrow><mn>2</mn></msup><mo>}</mo></mrow></mrow><mo>-</mo><msqrt><mrow><mi>E</mi><mo>⁢</mo><mrow><mo>{</mo><msup><mrow><mo></mo><msub><mi>S</mi><mi>k</mi></msub><mo></mo></mrow><mn>4</mn></msup><mo>}</mo></mrow></mrow></msqrt></mrow><mn>4</mn></mroot></mrow></mtd><mtd><mrow><mo>(</mo><mn>16</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0011.tif" /><br /> and the noise can be estimated as: <maths id="MATH-US-00012" num="00012"><math overflow="scroll"><mtable><mtr><mtd><mrow><msubsup><mover><mi>σ</mi><mo>^</mo></mover><mi>k</mi><mn>2</mn></msubsup><mo>=</mo><mrow><msqrt><mrow><mi>E</mi><mo>⁢</mo><mrow><mo>{</mo><msup><mrow><mo></mo><msub><mi>S</mi><mi>k</mi></msub><mo></mo></mrow><mn>2</mn></msup><mo>}</mo></mrow></mrow></msqrt><mo>-</mo><msubsup><mover><mi>a</mi><mo>^</mo></mover><mi>k</mi><mn>2</mn></msubsup></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>17</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0012.tif" /><br /> The estimates of equations (16) and (17) are inserted into equation (14) to obtain the weight.

0018Simulations were performed using adaptive weighting as described in equations (14), (16), and (17). Although long-term estimates without fading yielded good results, a compromise must be reached between long filter time constants for accurate estimation versus short filter time constants needed to track varying statistics due to fading.

0019In the Digital Audio Broadcasting (DAB) simulation, the OFDM symbol rate of 689.0625 Hz was chosen with a fading bandwidth of 13 Hz. Then the reciprocal of the fading bandwidth is about 53 symbols in this case. A filter time constant of 16 symbols was chosen since this time constant must be small compared to the fading time. Unfortunately, the statistical estimation errors over this short filter time yielded poor performance results for the adaptive weighting compared to what would be possible with perfect statistical estimation. Even reducing the fading bandwidth down to 3 Hz and increasing the filter time constant to 64 samples left a significant loss.

0020Equations (16) and (17) reveal that, in effect, quantities raised to the fourth power are subtracted to yield smaller numbers. This situation is most pronounced when the signal and noise powers are approximately equal, resulting in large estimation errors. Simulation results support this observation. Therefore another estimator is sought that does not rely upon subtraction fourth order statistics. The desired estimation technique should be designed to accommodate a fading bandwidth of up to 13 Hz for maximum vehicle speeds in the FM band around 100 MHz.

0021The optimum soft-symbol weight to be applied before differential detection of QPSK can be described as a function of time (k index) and OFDM sub-carrier (n index). Similar to equation (14), this weight is: <maths id="MATH-US-00013" num="00013"><math overflow="scroll"><mtable><mtr><mtd><mrow><msub><mi>w</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>=</mo><mfrac><msub><mi>a</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><msqrt><mrow><mrow><mn>2</mn><mo>⁢</mo><msubsup><mi>a</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow><mn>2</mn></msubsup><mo>⁢</mo><msubsup><mi>σ</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow><mn>2</mn></msubsup></mrow><mo>+</mo><msubsup><mi>σ</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow><mn>4</mn></msubsup></mrow></msqrt></mfrac></mrow></mtd><mtd><mrow><mo>(</mo><mn>18</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0013.tif" /><br /> where a_k,nis the fading coefficient of the k^thsymbol for the n^thsub-carrier, and σ_k,nis the corresponding standard deviation of the noise or interference, both prior to differential detection. Notice that the weight of equation (18) is the square root of equation (14). This is a result of the reasonable assumption that the weight changes slowly over the symbol-pair time used in the differential detection. In effect, the differential detection squares the predetection weight of equation (18), which would result in equation (14). A method for improving the statistical estimates of equation (18) is sought.

0022Practical methods for estimating CSI and weights using pre-differentially detected soft-symbols and weight also applied to the soft decision symbol prior to differential detection will now be discussed.

0023For moderate to high SNR, the weight of equation (18) can be conveniently approximated by: <maths id="MATH-US-00014" num="00014"><math overflow="scroll"><mtable><mtr><mtd><mrow><mrow><mrow><munder><mi>lim</mi><mrow><mi>SNR</mi><mo>→</mo><mi>∞</mi></mrow></munder><mo>⁢</mo><msub><mi>w</mi><mi>k</mi></msub></mrow><mo>=</mo><mfrac><mn>1</mn><msqrt><mrow><mn>2</mn><mo>⁢</mo><mstyle><mtext> </mtext></mstyle><mo>⁢</mo><msubsup><mi>σ</mi><mi>k</mi><mn>2</mn></msubsup></mrow></msqrt></mfrac></mrow><mo>,</mo></mrow></mtd><mtd><mrow><mo>(</mo><mn>19</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0014.tif" /><br /> where simple statistical measurements were used to estimate σ². However, simulation confirmed that this weight estimate performed poorly during times when the SNR was very low due to fading interference. For example, the optimum weight would have suppressed the noisy samples more than the high SNR approximation to the weight. Therefore, another approximation was sought which would estimate CSI statistics over a large SNR range. Furthermore the estimate should not be sensitive to a gaussian noise or interference assumption, and should be estimated with sufficient accuracy in a time (filter time constant) significantly less than the reciprocal of the fading bandwidth.

0024A simple and robust estimation technique evolved after simulation and some experimentation. This estimation technique approximates the previously-defined weight expressions, but uses lower-order statistical approximations. This technique is described in the following 4 steps.

00001. Create a sequence v_k,nfor each QPSK sub-carrier consisting of the magnitudes of the complex soft decision outputs s_k,nfrom the matched filter for the n^thsub-carrier. <br /><i>v</i>_k,n<i>=|s</i>_k,n|  (20)<br /> 2. Create a sequence d_k,nconsisting of the differences of successive time samples of v_k,n. <br /><i>d</i>_k,n<i>=v</i>_k,n<i>−v</i>_k−1,n  (21)<br /> 3. Filter the sequences v_k,nand d_k,nusing second-order digital IIR filters, then compensate for any differences in effective group delay to yield sequences filtv_k,nand filtd_k,n. The time constant for the filtv_k,nfilter should be somewhat smaller than the reciprocal of the fading bandwidth, while the time constant for the filtd_k,nfilter can be somewhat larger. These sequences are representative (approximately proportional) of the local mean and standard deviation of the sequence v_k,n. <br /> 4. The sequence of weights for the soft decisions for each sub-carrier to be applied prior to differential detection is defined as <maths id="MATH-US-00015" num="00015"><math overflow="scroll"><mtable><mtr><mtd><mrow><msub><mi>w</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>=</mo><mfrac><mn>1</mn><mrow><msub><mi>filtd</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>·</mo><mrow><mo>(</mo><mrow><mn>1</mn><mo>+</mo><msup><mrow><mo>(</mo><mfrac><msub><mi>filtd</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mrow><msub><mi>filtv</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>-</mo><msub><mi>filtd</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub></mrow></mfrac><mo>)</mo></mrow><mn>4</mn></msup></mrow><mo>)</mo></mrow></mrow></mfrac></mrow></mtd><mtd><mrow><mo>(</mo><mn>22</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0015.tif" /><br /> To prevent numerical overflow, check to ensure that filtv_k,n>1.5·filtd_k,nin equation (22); otherwise, set the weight to zero. Simulation results verified that this weight yields good performance under a variety of channel impairments with fading and interference.

0025The values of filtd_k,nand filtv_k,nare estimated using filtering techniques described next. Filtering is performed first for each sub-carrier at the k^thsymbol instant in time. Then the rows of filtd_k,nand filtv_k,nare simply updated across the N sub-carriers. Equation (23) filters the sequences v_k,nwith a time delay of approximately 16 symbols, and equation (24) filters the sequences d_k,nwith a time delay of approximately 64 symbols. Both filters have a zero frequency gain of nearly unity. <maths id="MATH-US-00016" num="00016"><math overflow="scroll"><mtable><mtr><mtd><mrow><msub><mi>subv</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>=</mo><mfrac><mrow><mrow><mn>960</mn><mo>·</mo><msub><mi>subv</mi><mrow><mrow><mi>k</mi><mo>-</mo><mn>1</mn></mrow><mo>,</mo><mi>n</mi></mrow></msub></mrow><mo>-</mo><mrow><mn>451</mn><mo>·</mo><msub><mi>subv</mi><mrow><mrow><mi>k</mi><mo>-</mo><mn>2</mn></mrow><mo>,</mo><mi>n</mi></mrow></msub></mrow><mo>+</mo><mrow><mn>3</mn><mo>·</mo><msub><mi>v</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub></mrow></mrow><mn>512</mn></mfrac></mrow></mtd><mtd><mrow><mo>(</mo><mn>23</mn><mo>)</mo></mrow></mtd></mtr><mtr><mtd><mrow><msub><mi>subd</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>=</mo><mfrac><mrow><mrow><mn>16128</mn><mo>·</mo><msub><mi>subd</mi><mrow><mrow><mi>k</mi><mo>-</mo><mn>1</mn></mrow><mo>,</mo><mi>n</mi></mrow></msub></mrow><mo>-</mo><mrow><mn>7939</mn><mo>·</mo><msub><mi>subd</mi><mrow><mrow><mi>k</mi><mo>-</mo><mn>2</mn></mrow><mo>,</mo><mi>n</mi></mrow></msub></mrow><mo>+</mo><mrow><mn>3</mn><mo>·</mo><msub><mi>d</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub></mrow></mrow><mn>8192</mn></mfrac></mrow></mtd><mtd><mrow><mo>(</mo><mn>24</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0016.tif" />

0026Additional filtering is performed across the N sub-carriers. Smoothing the estimates across the N sub-carriers requires 3 passes of a simple IIR filter. The first pass sets the appropriate initial condition of the filter, but does not update the estimates. The direction of the second pass is reversed from the first, while the third pass is reversed again. This results in an approximately symmetric (linear phase) filter characteristic which is desirable for providing the estimates on the center carrier. Although it is impossible to provide this symmetric filtering for the sub-carriers at each end of the band, the impulse response “tails” are folded back into the active sub-carriers.

0027The first pass across the sub-carriers sets the initial values of filtv_N−1and filtd_N−1without replacing the time-filtered values for each sub-carrier. The time index k is ignored here since it is understood that the filtering over the sub-carriers is performed over each k^thOFDM symbol.

0000ti <i>filtv</i>_N−1(1−β)·<i>filtv</i>_N−1<i>+β·subv</i>_n; <br /><i>filtd</i>_N−1(1−β)·<i>filtd</i>_N−1<i>+β·subd</i>_n  (25)<br /><i>n=</i>0,1<i>, . . . N−</i>1

0028The second pass smoothes the values across the filtered estimates for each sub-carrier, subv and subd. <br /><i>filtv</i>_n(1−β)·<i>filtv</i>_n+1<i>+β·subv</i>_n;<br /><i>filtd</i>_n(1−β)·<i>filtd</i>_n+1<i>+β·subd</i>_n;  (26)<br /><i>n=N−</i>2<i>,N−</i>3, . . . 0

0029The third pass smoothes the frequency values again to achieve a nearly symmetrical impulse response (except for the sub-carriers near the endpoints). <br /><i>filtv</i>_n(1−β)·<i>filtv</i>_n−1<i>+β·filtv</i>_n;<br /><i>filtd</i>_n(1−β)·<i>filtd</i>_n−1<i>+β·filtd</i>_n;  (27)<br /><i>n=</i>1,2<i>, . . . N−</i>1.<br /> The resulting filtered values for filtv and filtd computed in equations (26) and (27) are used in equation (22) at each OFDM symbol time to yield the appropriate weight for each soft symbol prior to differential detection, but after matched filtering, in the receiver.

0030<figref idref="DRAWINGS">FIG. 1</figref> is a diagram showing the steps of the method of this invention. Block <b>10</b> shows the step of receiving a plurality of sub-carriers modulated by digital information. The sub-carriers are filtered as shown in block <b>12</b> to produce complex soft decision outputs. These outputs are used in block <b>14</b> to create a first sequence of the magnitudes of the complex soft decision outputs. Block <b>16</b> shows that the differences between successive samples in the first sequence are determined. A second sequence of the differences between successive samples in the first sequence is created as shown in block <b>18</b>. The first and second sequences are filtered as in block <b>20</b> and compensated for differences in effective group delay of the first and second sequences as in block <b>22</b> to produce third and fourth sequences in block <b>24</b>. The third and fourth sequences are used to determine a plurality of weights as shown in block <b>26</b>, and the plurality of weights are applied to the complex soft decision outputs as shown in block <b>28</b>.

0031The above discussion relates to differentially detected QPSK. The derivation for the weight using coherent detection of QPSK or (BPSK) would be similar to the technique discussed above for differentially detected QPSK. The only modification should be in the expression presented in equation (22). It can be predicted that the weight should be of the form: <maths id="MATH-US-00017" num="00017"><math overflow="scroll"><mtable><mtr><mtd><mrow><msub><mi>w</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>=</mo><mfrac><mrow><msub><mi>filtv</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>-</mo><mrow><mi>c</mi><mo>·</mo><msub><mi>filtd</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub></mrow></mrow><msubsup><mi>filtd</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow><mn>2</mn></msubsup></mfrac></mrow></mtd><mtd><mrow><mo>(</mo><mn>27</mn><mo>)</mo></mrow></mtd></mtr></mtable></math></maths><img file="US6895060B2_D0017.tif" /><br /> where c is a constant to be defined by empirical methods.

0032Techniques for estimating the optimum soft-decision weight for QPSK symbols prior to Viterbi decoding were described. These techniques apply to coherent and differential detection of single or multi-carrier (OFDM) QPSK, with and without multipath fading of the signal of interest or the interferer. The fading cases can necessitate a compromise between accurate CSI estimation and agility of the CSI to track the fading signal or noise components.

0033The application of soft-decision weighting for an OFDM system with independently faded signal and noise (interference) can improve BER performance. This weight is applied to the soft symbols prior to deinterleaving and Viterbi decoding. The optimal weight for each soft symbol over time and sub-carriers is estimated through an expression using filtered statistical estimates of channel state information (CSI) about the signal and noise components of the received symbols. The above CSI estimation and weighting techniques can apply to both coherent and differentially detected symbols, and to single and multi-carrier (OFDM) modulation in the presence of multipath fading and colored noise statistics.

0034<figref idref="DRAWINGS">FIG. 2</figref> is a functional block diagram showing the mapping and processing of bits through a portion of an FM receiver that operates in accordance with the method of the invention. A plurality of OFDM carriers <b>30</b> are received and converted to bit streams on lines <b>32</b> by receiver circuit <b>34</b>. Circuit <b>34</b> includes a digitizer, carrier synchronization, symbol synchronization, and matched filters all operating in accordance with well known techniques to produce the bit streams on lines <b>36</b>. Block <b>38</b> represents the channel state estimates and weighting processes performed according to <figref idref="DRAWINGS">FIG. 1</figref> to produce weighted bit streams on lines <b>40</b>. Block <b>42</b> shows that the bit streams are deallocated from the carriers and delivered to a deinterleaver <b>44</b>. The output of the deinterleaver is multiplexed to a single bit stream and passed to a Viterbi decoder <b>46</b> decodes the single bit stream. Soft-decision Viterbi decoding with (near) optimum soft-decision weighting for maximum ratio combining (MRC) for differentially detected QPSK sub-carrier symbols is employed to minimize losses over the channel. The output of the Viterbi decoder is subject to additional signal processing, which is not part of the present invention, and passed to output stages, as illustrated by block <b>48</b>, to produce the desired output from the receiver.

0035While the present invention has been described in terms of what are at present believed to be its preferred embodiments, it will be apparent to those skilled in the art that various changes may be made to the embodiments described above without departing from the scope of the invention as set forth in the following claims.